Although a section in the bibliography is devoted to the refractivity of air, it's buried down near the end of the monster file. Besides, a little more detailed discussion of the dispersion formulae for air seems to be called for.
The main problem is that the refractivity of air is difficult to measure accurately, so that there have been many re-measurements, and several different formulae have been used to represent the dispersion curve of air by different authors. Worse yet, the older formulae, which have long been known to be incorrect, have become fossilized in handbooks, and copied by more recent handbooks from the older ones, so that obsolete and inaccurate formulae are often cited and used.
Note that the refractivity is usually given for 15° C, and refers to dry air, not air with some water vapor in it. The argument is usually vacuum wavenumber, not just the reciprocal of the wavelength in air. Sometimes the formula given refers to air free of CO2 as well as water vapor. Because the refractivity, (n−1), is generally less than 3×10−4, it's quite common to have the formula give either 106 or 108 times the actual refractivity. Bear all these facts in mind when using any refractivity formula. Read all the fine print before using any formula!
Finally, there is more than one “refractive index” of interest, and the right one to use depends on the kind of measurement being made. For angular refraction (which is what I deal with on these Web pages), the appropriate index is the “phase index,” which is the one usually meant if no such distinction is made. However, for radar and laser geodimetry, where an electromagnetic signal is modulated, it is usually the group velocity rather than the phase velocity that is of interest; then the “group index” is required. If observations are made in or near absorption lines, things become still more complex. Fortunately, these niceties are mainly of concern for the geodesists, and will be ignored here.
In its day, the discussion by
H. Barrell, J. E. Sears
The Refraction and Dispersion of Air for the Visible Spectrum
Phil. Trans. Roy. Soc. A , 238, (1939)
was quite influential. Its numerical constants are now quite obsolete; its refractivity error approaches a part in 1000 near the atmospheric transmission cutoff just below 300 nm, and is about 1 part in 5000 or 6000 through most of the visible spectrum. (However, the discussion of the physical effects that must be taken into account is still worth reading, even though the numerical values have been superseded.)
Unfortunately, one occasionally still finds papers that use it. (The tip-off that this obsolete formula is being used is that it has the form of a 2-term Cauchy dispersion formula: a constant, a term in λ−2, and a term in λ−4.)
An example is the paper by J. Gubler and D. Tytler in PASP 110, 738 (1998). They got the B&S formula from A. T. Sinclair's NAO Technical Note (1982), which doesn't sound too obsolete. But Sinclair got it from the IAG's Bulletin Géodésique (1963), p. 390 — which of course got it from Barrell & Sears (1939). By the time Gubler & Tytler picked it up, the formula was out of date by almost 60 years!
Unfortunately, this incorrect and obsolete formula was also used by Hohenkerk & Sinclair (1985), from which it was copied in the revised (1992) edition of the Explanatory Supplement (see the middle of p. 142 for the obsolete formula there); so we can expect to continue to see it crop up again in the future. And sure enough, it's still there in the 2013 third edition. Obviously, astronomers don't care about refraction any more.
B. Edlén
Dispersion of standard air
J. O. S. A. 43, 339–344 (1953)
This is the formula with a 41 in the denominator of the last term. It's quoted on p. 119 of the 2nd Edition of Allen's Astrophysical Quantities (1963), and on p. 124 of the 3rd Edition (1973). (A modified form of this formula was also included in the IAG's Bulletin Géodésique (1963), p. 390. Recently, the IAG finally got clued in and dumped these obsolete formulae.) Though it's still often cited, it's known to be wrong. Don't use it!
It's inexcusable that it was still given as “the” refractive index of air on p. 262 of the 4th edition of Allen's Astrophysical Quantities in the year 2000 — 36 years after it was repudiated by Edlén himself! (Worse yet, this handbook even misprints it with an egregious typo that reduces the formula to rubbish, on p. 257. They're charging $99 for this ?)
B. Edlén
The refractive index of air
Metrologia 2, 71–80 (1966)
J. C. Owens
Optical refractive index of air: dependence on pressure, temperature and composition
Appl. Opt. 6, 51–59 (1967)
Owens, like Barrell & Sears, separated out the contribution from carbon dioxide. He claimed an accuracy of 1 part in 109 for the refractive index, or 3 parts in 106 for the refractivity. Like Barrell & Sears, Owens considered compressibility effects (i.e., deviations from the ideal-gas law); he found them to be about 10−4 at 245 K, increasing rapidly at lower T. (From his figure, it looks like 2×10−4 at 200K).
Owens showed that Edlén (1966) is off by 7×10−8 at −30°C, and by 4×10−7 at 45°C, 100% RH. Unfortunately, the 1966 Edlén formula is still widely quoted and used.
E. R. Peck, K. Reeder
Dispersion of air
JOSA 62, 958–962 (1972)
made new measurements in the infrared. They showed that Edlén was certainly wrong in the IR, and that one could obtain accuracy as good as the data (except at wavelengths below 0.23 µm) with a 2-term Sellmeier dispersion formula, which involves only 4 parameters instead of 5. This accuracy is a little better than 2 parts in 109 for the refractive index, from 0.23µm in the UV to 1.69µm in the near IR.
Personally, I like to use their simple formula for calculating atmospheric refraction. It's “good enough” for most purposes, as long as you don't need to worry about water vapor, or the middle infrared. (I have a plot of their dispersion curve on another page.)
F. E. Jones
The refractivity of air
J. Res. NBS 86, 27–32 (1981)
reconsidered the real-gas effects, and performed a careful error analysis:
The major contributors to the uncertainty in refractivity are the uncertainties in the measurements of temperature and pressure. The magnitude of the uncertainty due to variation in CO2 concentration can approach that of the uncertainties due to the pressure and temperature measurements. Therefore, the CO2 concentration should be treated as a variable and should be observed.
This statement received excellent confirmation a decade later in the 1993 and 1994 papers of Birch and Downs (see below), who measured the effects of varying CO2 concentration in laboratory air.
Unfortunately, Jones seems to have overlooked the Peck – Reeder work. He produced a formula good for the visible region with an error in the refractivity of a few parts in 108.
H. Matsumoto
The refractive index of moist air in the 3-µm region
Metrologia 18, 49–52 (1982)
measured the effects of water vapor, and found considerable effects due to its near-IR absorptions.
Then additional errors due to water vapor, even in the visible, were found by
K. P. Birch, M. J. Downs
The results of a comparison between calculated and measured values of the refractive index of air
J. Phys. E: Sci. Instrum. 21, 694–695 (1988)
and confirmed by
J. Beers, T. Doiron
Verification of revised water vapour correction to the refractive index of air
Metrologia 29, 315–316 (1992)
P. E. Ciddor
Refractive index of air: new equations for the visible and near infrared
Appl. Opt. 35, 1566–1573 (1996)
We can regard Ciddor's results as definitive, at least for the present. They have been adopted as the basis of a new standard by the International Association of Geodesy (IAG). Geodetic applications often require the group rather than the phase index; see
P. E. Ciddor, R. J. Hill
Refractive Index of Air. 2. Group Index.
Applied Optics (Lasers, Photonics and Environmental Optics) , 38, 1663–1667 (1999)
Unfortunately, there is a sign error in Eq. (B2) of Ciddor and Hill's paper; see the correction in
F. Pollinger
Refractive index of air. 2. Group index: comment
Applied Optics 59, No. 31, pp. 9771-9772 (2020)
Soon after Ciddor's first paper, a new set of highly accurate measurements at 4 wavelengths in the visible spectrum was published by
G. Bönsch and E. PotulskiTheir work focuses primarily on determining the effects of water vapor and CO2 content, and takes account of a slight adjustment between older practical temperature scales and the ITS-90 scale, which is almost a hundredth of a degree.
Measurement of the refractive index of air and comparison with modified Edlén's formulae
Metrologia 35, 133–139 (1998)
They start with the “new” Edlén formula of 1966. The aim is to provide the highest possible accuracy for laboratory conditions, so they consider only a small temperature range about 20°C and a CO2 mixing ratio near 400 ppm, where linear formulae are adequate; their results are not sufficiently general for field use in the actual atmosphere. They give a nice practical summary of their results in an Appendix.
As they essentially confirm the results of previous recent studies — in particular, the water-vapor corrections of Birch & Downs and more recent workers — I would regard their paper as indirectly supporting Ciddor's results. The uncertainty in the refractive index of air is about 1 unit in the 8th decimal place.
For further discussion of these matters, see the Web pages of NIST and the IAG. NIST also has a useful Web page at https://www.nist.gov/publications/index-refraction-air that provides practical information about the refractivity of air.
The above discussion refers to the effects of water vapor in the visible spectrum. But water vapor also has strong absorptions in the near infrared, which (by classical dispersion theory) produce considerable effects on the dispersion curve there. The following reference gives an example of the size of these effects out to 25 microns (in spite of the narrower scope suggested by its title):
R. J. Mathar
Calculated refractivity of water vapor and moist air in the atmospheric window at 10 μm
Appl. Opt. 43, 928–932 (2004)
Similar calculations were published by
at about the same time; and Mathar subsequently published convenient series expansions for the infrared refractivity of air, given the humidity:M. M. Colavita, M. R. Swain, R. L. Akeson, C. D. Koresko, and R. J. Hill
Effects of atmospheric water vapor on infrared interferometry
Pub. Astron. Soc. Pacific 116, 876–885 (2004)
R. J. Mathar
Refractive index of humid air in the infrared: Model fits
J. Optics A: Pure Appl. Opt. 9, 470–476 (2007)
These authors point out that the large decrease in refractivity in the middle IR — in particular, in the important 10μm window — is due to the strong absorptions of the CO2 fundamental near 15μm and the pure-rotational spectrum of water vapor at longer wavelengths (which Colavita et al. repeatedly, and erroneously, call a vibration-rotation band).
andK. P. Birch, M. J. Downs
An updated Edlén equation for the refractive index of air
Metrologia 30, 155–162 (1993)
K. P. Birch, M. J. Downs
Correction to the updated Edlén equation for the refractive index of air
Metrologia 31, 315–316 (1994)
As their titles indicate, these papers were intended to revive the (new) Edlén formula. Although the first paper contains a section headed “Review of Developments Since 1966”, it ignores the works of Owens (1967), Peck & Reeder (1972), and Jones (1981). [Well, there's a reference to the 1967 paper by “Owen” (sic), but only to cite his percentage composition for standard air; there's no mention of his work on the dispersion relation.]
Their 1993 paper discusses a variety of effects: water vapor, compressibility (i.e., real-gas effects in the equation of state), revisions of the temperature scale, and CO2 content. They retained Edlén's dispersion formula, and made measurements only at the He-Ne laser wavelength (633 nm). Unfortunately, as the 1994 paper points out, they put the temperature-scale revision in the wrong way, so that the final results are wrong. (Table 1 of the 1994 paper is a corrected version of Table 4 in the 1993 paper.)
However, the main value of this work was to show convincingly that the CO2 content of laboratory air is usually appreciably higher than in the free atmosphere — I suppose because laboratories are full of people continually exhaling the gas — and that the effect of this on the refractivity is plainly measurable. In this sense, I take their work as experimental confirmation of the work of Jones (1981).
A proper treatment of the CO2 problem would take account of the fact that the increase in its abundance is at the expense of molecular oxygen. The dispersion curves of both gases should be treated accurately, instead of simply scaling the refractivity of air, as Birch & Downs do. In any case, their results are by now completely superseded by Ciddor's.
Thanks to William M. Robertson for providing copies of these Birch & Downs papers to me.
A. Picard, R. S. Davis, M. Gläser, and K. Fujii
Revised formula for the density of moist air (CIPM-2007)
Metrologia 45, 149–155 (2008)
This paper does not discuss refractivity, but deals only with the actual density of air. Its main contribution is to incorporate more recent values for the argon abundance in air, and to employ a realistic equation of state. These tiny corrections — a few parts in 105 — are smaller than the required accuracy for refraction models.
Similarly, a change of 0.1 mm of mercury changes the density and hence the refractivity of air by about a part in 104, so that still smaller pressure changes might be significant. This is also the level at which deviations from the ideal gas law become appreciable.
Stone says that, for accurate positional astrometry, errors should be kept below 0.05 arc seconds at zenith distances up to 70°, where the sea-level refraction is about 147 arc seconds. This refraction is nearly 3000 times the allowed error.
For most purposes, we can neglect effects smaller than 1 or 2 seconds of arc at the horizon, corresponding to relative errors in density of 1 part in 1000, or perhaps 1 part in 2000. It appears that an error of 1 part in a few thousand is small enough for astronomers (though not for geodesists, geophysicists, and some surveyors).
This means an accuracy of 1 part in 3000 for the refractivity, or 1 part in 107 for the refractive index, is adequate for most astronomical purposes. That would make the 1966 Edlén formula acceptable. But, as the more recent formulae — in particular, the Peck – Reeder formula — are more reliable, as well as less expensive to calculate, I think one of them should be used.
Note that 1 part in 3000 requires measuring temperatures to 0.1°C or better accuracy (not just precision); pressures must be accurate to about 1/3 of a millibar; and the water-vapor mixing ratio should be known to a relative accuracy of about 10%.
For astronomical refraction calculations in the visible spectrum, the Peck – Reeder formula seems a good compromise between accuracy and economy of calculation. I recommend it.
For other purposes, such as geodesy, more accuracy is certainly required. Then one should adopt Ciddor's results in the visible, and take account of the large dispersive effects of water vapor and CO2 in the infrared.
In the real world, the refractivity of air is not the main problem. Deviations of the idealized refraction models from the actual physical situation are the real limitations. These are of two main kinds: systematic errors in measuring temperature and pressure; and deviations from the usual assumption of spherical symmetry. The latter are discussed in various items in the annotated bibliography whose comments contain the word “tilt”. (See the special section of the bibliograpjhy headed TILT FILE.)
The effects of tilt are significant for typical tilts of the isopycnic surfaces, which are about a minute of arc. This introduces a zero-point error in refractions that can be a few hundredths of a second of arc at the zenith. The error depends on the difference between the azimuth of a star and the direction of the wind; it increases with the square of the secant of the zenith distance. This is significant for geophysics and geodesy, as it is a considerable fraction of the Chandler wobble, and is significant in measurements of the deflection of the vertical, which are important for determining the shape of the geoid, the viscosity of the liquid core, solid-body tides, and other geophysical problems. Near the horizon, the effect of tilt can exceed a minute of arc, and produces changes in geographic positions inferred from the Global Navigational Satellite System that can exceed a meter.
For a more thorough and authoritative discussion of the refractivity of air that takes account of the effects of both humidity and CO2 variations, see the NIST web-page:
http://emtoolbox.nist.gov/Wavelength/Documentation.asp#EdlenorCiddor
Copyright © 2003 – 2007, 2010, 2011, 2017 – 2021, 2024 Andrew T. Young
or the
GF home page
or the Overview page